Repository navigation
[WIP] Fix stale routed expert accumulation across MoE forwards - #3482
Draft
cla7aye15I4nd wants to merge 1 commit into
Draft
cla7aye15I4nd wants to merge 1 commit into
cla7aye15I4nd wants to merge 1 commit into
Conversation
|
👋 Hi! Thank you for contributing to the TileLang project. Please remember to run We appreciate you taking this step! Our team will review your contribution, and we look forward to your awesome work! 🚀 |
Contributor
|
Important Draft PR not reviewedDraft PRs are not automatically reviewed by default.
To automatically review draft PRs, update your CodeRabbit configuration: reviews:
auto_review:
drafts: true
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
This branch has not been deployed
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Reusing one
MoEinstance adds routed expert results from previous forwards becausescatter_reduceincludes the persistent destination buffer. With valid inputs, the first call matches the PyTorch reference but the second call is wrong.Clear
expert_cachein the routed stream immediately before reduction. Add a regression that reuses the same module for identical and changed inputs with top-k 1 and 2.Validation:
Run with
TILELANG_EXECUTION_BACKEND=nvrtc pytest examples/fusedmoe/test_repeated_forward.py examples/fusedmoe/test_example_fusedmoe.py -q.GPU validation used an RTX 5070, Python 3.12, PyTorch 2.14.1+cu130, and the TileLang 0.1.15 wheel running the changed example files from base
7a8ded9f1d91b0693fa82467656311947cfd8922. This was not a full source build or an all-backend test. NVCC host compilation was blocked by a CUDA/glibc header incompatibility; NVRTC compiled and ran the GPU kernels. Concurrent forwards on one instance and performance were not evaluated.